Below are several hearing loss compensation systems with corresponding audio examples across different signal-to-noise ratios (SNRs). The patient has a moderately, sloping hearing loss. Here’s a brief explanation of each system:
Unprocessed : The original noisy audio without any processing.
HLC (audiogram divided by 2): This system is derived from the audiogram divided by 2 in the log domain. Optimized on the auditory-nerve output of the Zilany model.
HLC (audiogram without division) : This system uses the full audiogram without division in the log domain. Optimized on the auditory-nerve output of the Zilany model.
NR_NAL_R: This system applies noise reduction followed by NAL-R filtering.
HLC-NR (audiogram divided by 2): This system performs joint HLC and NR and is derived from the the audiogram divided by 2 in the log domain. Optimized on the auditory-nerve output of the Zilany model.
DNN-HA: Deep Neural Network Hearing Aid system. More details available at DNN-HA GitHub.
HLC (IC, audiogram divided by 2): Similar to the audiogram divided by 2 system, but appended with Inferior Colliculus (IC) modeling.
HLC (IC, audiogram not divided by 2): Similar to the no division, but appended with Inferior Colliculus (IC) modeling.
HLC-NR (IC, audiogram divided by 2): Similar to the HLC-NR, auditory nerve model appended with Inferior Colliculus (IC) modeling. Importantly, this system uses Wave-U-Net (WUN).